An Attention-Based Fusion of ResNet50 and InceptionV3 Model for Water Meter Digit Recognition

Abstract

Digital water meter digit recognition from images of water meter readings is a challenging research problem. One key reason is that this might be a lack of publicly available datasets to develop such methods. Another reason is the digits suffer from poor quality. In this work, we develop a dataset, called MR-AMR-v1, which comprises 10 different digits (0 to 9) that are commonly found in electrical and electronic water meter readings. Additionally, we generate a synthetic benchmarking dataset to make the proposed model robust. We propose a weighted probability averaging ensemble-based water meter digit recognition method applied to snapshots of the Fourier transformed convolution block attention module (FCBAM) aided combined ResNet50-InceptionV3 architecture. This benchmarking method achieves an accuracy of 88% on test set images (benchmarking data). Our model also achieves a high accuracy of 97.73% on the MNIST dataset. We benchmark the result on this dataset using the proposed method after performing an exhaustive set of experiments.

Received: 10 June 2023 | Revised: 4 September 2023 | Accepted: 10 October 2023

Conflicts of Interest

Palaiahnakote Shivakumara is an editor-in-chief for Artificial Intelligence and Applications, and was not involved in the editorial review or the decision to publish this article. The authors declare that they have no conflicts of interest to this work. 

Data Availability Statement

The main MR-AMR dataset is available at https://data mendeley.com/datasets/8xjhrrk9rx and the version-2 (which includes the challenge benchmarking dataset) is available at https://www.kaggle.com/datasets/ayush02102001/watermeter-data-recognition.

Author Contribution Statement

Ayush Roy: Conceptualization, Methodology, Software, Investigation, Writing - original draft, Visualization. P. Shivakumara: Conceptualization, Validation, Writing - original draft, Supervision, Project administration. Umapada Pal: Writing - review & editing.